Space Assembly and Service via Self-Reconfiguration

1
Space Assembly and Service via Self-Reconfiguratio
n

• Wei-Min Shen and Peter WillUSC/ISI Polymorphic
  Robotics Laboratory
• Berok Khoshnevis
• USC Industrial and Systems Engineering Department
  
• George Bekey
• USC Computer Science Department
• Space Solar Power Concept Technology Maturation
  Program (SCTM) Technical Interchange Meeting
• NASA Glenn Research Center, Cleveland Ohio

2
ISI Polymorphic Robotics Labhttp//www.isi.edu/ro
bots

• Mission
• To build Self-Reconfigurable Systems such as
  metamorphic robots, agents, and smart structures
  that go where biological systems have not gone
  before!!!
• Projects and Awards
• YODA (1996) The 2nd place in AAAI competition
• Dreamteam (1997) RoboCup World Champion
• Intelligent Motion Surface in MEMS (1996-98)
• CONRO Self-Reconfigurable Robots (1998-)
• People, Robots and Facilities
• Experienced and talented research team
• 3 Denny robots, 5 SoccerBots, 18 CONRO modls
• Large labs and workshops, many instrumentations

3
Outline

• Motivation for Self-Assembly in Space
• Three Enabling Technologies
• Based on Self-Reconfigurable Robots
• Proposed Evaluation Experiments
• Research Plan for SSPS
• Future Directions

4
Motivation for Self-Assembly

• Cost Effective
• For a 10KM SSPS
• gt2,500 hours of astronaut space walk
• 4/11/2002, girder assembly (26 hours)
• gt3 billion for assembly cost
• Feasible Strategy
• Most jobs by self-assembly
• Critical jobs done by astronauts

5
A Vision for Space Self-Assembly
6
Three Enabling Technologies

• Intelligent and Reconfigurable Component (IRC)
• Can free-float and dock to form structures
• Free-flying Fiber Match-Maker Robots (FIMER)
• Can search, navigate, bring-together and dock
  IRCs
• Distributed Process Controller (DPC)
• Can plan self-assembly in a distributed manner
  and recover from unexpected situations

7
Self-Reconfigurable Robots
8
CONRO Self-Reconfigurable Modules
A network of physically coupled
agentsSelf-assembling into various
configurations!
9
Live Surgery Reconfiguration
10
Beyond-Bio Self-Reconfiguration
11
Challenges in Control

• Distributed
• Autonomous modules must be coordinated by local
  configuration information (no unique IDs or brain
  modules)
• Dynamic
• Network and configuration topology changes
• Asynchronous
• Communication with no real-time clocks, global or
  local
• Scalable
• Weak local actions vs. grand global effects
• Fault-tolerant
• Miniature and self-sufficient

12
Related Work

• Control approaches
• Control tables (Yim94)
• Multi-agents (Hogg2000)
• Finite State Machine (Rus2000)
• Decentralized and autonomous system (Mori84)
• Homeostatic control for resource allocation
  (Arkin88)
• Dynamic topology network (SiLin2000)

• Self-Reconfigurable robots
• Diffusion-reaction (Turing 52)
• Cebots (Fukuda Nakagawa90)
• Polybots (Yim 94)
• Metrics (Chirikijan 98)
• 3D structures (Murata 98)
• Self repair (Murata 2000)
• Molecules (KotayRus 98)
• Feather formation (Chuong 98)
• Self-Transform (Dubowsky00)

13
Digital Hormones

• Content-based messages
• No addresses nor identifiers
• Have finite life time
• Trigger different actions at different sites
• Floating in a global medium
• Propagated, not broadcast
• Internal circulation, not external deposit
  (pheromones)
• Preserve local autonomy for individual sites
• Hormones can modify module behaviors (RNA)

14
Mechanical Cells (M-Cell)
15
M-Cell Organizations
A module
A Snake
A 6-leg insect
Communication between two separate structures
16
M-Cell Control Software
Local Decision Engine
From the global Hormone Medium
To the global Hormone Medium
Local Programs
Local Actuators Sensors
Local State Knowledge
17
Discovering Topology
18
The Uses of Digital Hormones

• Communication in dynamic network
• Cooperation among distributed autonomous modules
• Locomotion
• Reconfiguration
• Synchronization
• Global effects by weak local actions
• Conflict resolution (multi hormone management)
• Navigation
• Shape adaptation and self-repairing

19
Hormones for Caterpillar Move

• A simple one-pass hormone from head to tail
• Controls and synchronizes all motor actions
• Independent from the length of the snake

20
Reconfigure Insect ? Snake
21
Hormone Activities
Active hormones
Actions
LTS
Start the reconfiguration
RCT
, RCT
, RCT
, RCT
Legs are activated
1
2
3
4
TAR, RCT
, RCT
, RCT
The tail inhabits RCT, and leg1 determines RCT
2
3
4
1
ALT, RCT
, RCT
, RCT
The tail assimilates leg1 and then accepts new RCT
2
3
4
TAR, RCT
, RCT
The tail inhabits RCT, and leg3 determines RCT
2
4
3
ALT, RCT
, RCT
The tail assimilates leg3 and then accepts new RCT
2
4
TAR, RCT
The tail inhabits RCT, and leg4 determines RCT
2
4
ALT, RCT
The tail assimilates leg4 and then accepts new RCT
2
TAR
The tail inhabits RCT, and leg2 determines RCT
2
ALT
The tail assimilates leg2 and then accepts new RCT
Æ
End the reconfiguration
22
Autonomous Docking

• A great challenge for self-reconfiguration
• Require precise sensor guidance
• Demand precision movement
• Complex dynamics in micro-gravity environment

23
Intelligent Reconfigurable Components
An IRC has (1) a controller, (2) a set of named
connectors, (3) wireless communication, (4)
self-locating system, and (5) short-range sensors
for docking guidance
24
Reconfigurable Connectors
25
FIMER Robots
Two-headed fiber/ropeFree-flying head
(6DOF)Navigate and dock to the
connectorsRail-in fiber to bring parts
togetherSimple arms to assist dockOnboard power
or refuel capability
26
FIMER Dynamics and Control
Find relevant connectors based on their location
informationRailing in the fiber only when there
is no tension
Research Issues Dynamics of tethered objects
in zero-gravity environment Speed
control Collision control Prevent
tangling
27
The Global Process Control

• How do modules know when and where to connect?
• Advantages for distributed control
• Coordination of autonomous modules without fixed
  brain
• Support dynamic configuration topology
• Asynchronous communication without global clocks
• Scalable support growing structures
• Fault-tolerance
• Self-repairing capability
• Self-replanning for unexpected events

28
Proposed Process Control

• Assumptions
• Modules have unique identifiers
• Assembly sequence embedded in modules
• Procedures
• Activate self when receiving a call for its ID or
  type
• Call FIMER robots to assist docking (when
  activated)
• Activate the next connectors to be docked

29
Proposed Experiments

• Build modules for autonomous planning,
  navigation, docking
• 2D flight-test on an air hockey table
• Extensible to future 3D flight-test in
  micro-gravity environment

30
Research Time Table
Task Time
Computer Simulation 0-3 month
Building 2D flight modules/robots 0-12 month
Control framework and algorithms 6-24 month
Forming simple 2D structures 12-24 month